[LASCOT2012] Measure and Manage Flow

The fifth iteration of this talk, presented at Lean Agile Scotland 2012.

Please note that the content of this talk changes after each event, because I'm continuously improving it using the feedback and the needs of the audience.

Transcript

1. Measure and Manage Flow in Practice @ZsoltFabok http://zsoltfabok.com/ #lascot12 http://www.leanagilescotland.com/ by Zsolt Fabok 2012.09.21
5. #1 Too many open items
6. WIP point of observation Visualize the situation with Cumulative Flow Diagram solved
7. The Cumulative Flow Diagram Done Started Queued lead time cycle time WIP backlog time number of work items It offers more than just the WIP...
8. ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ Q D 3 4 1 2 The simplest way of collecting data:
9. CFD doesn’t say too much about the throughput A quick detour:
10. work item WIP lead time /* detour */ time
11. lead time work item throughput 2 WIP /* detour */
12. work item 2 2 /* detour */
13. Let’s add more people to the project so that “things speed up”! /* detour */
14. work item 2 2 3 coordination + communication cost /* detour */
15. Actually, “things slow down”, so it was not a good idea (solve the right problem instead - systems thinking). End of the detour.
16. #2 It takes too much time to deliver
17. lead time CFD was not much help here... ...because we didn’t know much about the nature of the lead time
18. Distribution of lead times days count 0 3 5 8 10 13 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 22 28 33 56 average median* *Calculation of medians is a popular technique in summary statistics and summarizing statistical data, since it is simple to understand and easy to calculate, while also giving a measure that is more robust in the presence of outlier values
19. time spent on implementation (hours) time spent waiting (hours) % 1 63 98 7 57 90 2 62 97 2 62 97 3 61 96 Some examples of work items with 8-day lead time
21. 95% of the lead time was spent on waiting
22. Distribution of lead times days count 0 3 5 8 10 13 15 1 4 7 10 13 16 33 Before average 0 3 5 8 10 13 15 1 3 5 7 9 11 13 22 count days After median
23. #3 Still too many open work items
24. How many times the item has been rejected
25. 0 4 8 11 15 31-32 33-34 35-36 37-38 Number of rejected work items count week
26. 0 4 8 11 15 31-32 33-34 35-36 37-38 39-40 42-43 44-45 Number of rejected work items count week
27. #4 Being predictable
28. Sales: “I want to know when the new features can hit the market!” Management: “I want to know how much it will cost me!”
29. All the work items we had so far (~20 work items) ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ v v v v v v v v
30. Categorizing them into three groups ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ v v v v v v v v S M L
31. The lead time distribution ~ ~ ~ ~ ~ ~ ~ ~ v v v v M 0 1 1 2 3 3 4 1 2 3 4 5 6 7 8 9 12 13 16 SLA days count Expired
32. The spent time distribution ~ ~ ~ ~ ~ ~ ~ ~ v v v v M 0 1 2 3 4 5 6 6 7 8 9 10 SLA hours count
34. #5 Forced improvement
35. #1 We decided that we would force ourselves to keep the SLA #2 Nothing changed. Still the same ratio
36. Final thoughts
37. “If you start measuring something you start optimizing it, and I know it's the wrong thing to optimize.” A quote which is worth considering http://paulgraham.com/swan.html
38. The key ideas 1. We develop software not models (value) 2. Demand first, supply second 3. Observe the system (lead time, throughput) 4. Start measuring, look back if necessary 5. Manage 6. Mind that data expires 7. Goto step 3.
39. Thank you very much for your attention! http://zsoltfabok.com/ @ZsoltFabok